Develop a multi-agent solution with Microsoft Agent Framework

Why it matters

A prebuilt sequential orchestration is the shortest path to a multi-agent pipeline, so you need to know what it gives you and what it leaves out.

What to practise

Create a Foundry project and model deployment from the Foundry Toolkit extension for VS Code, which some labels still call AI Toolkit. Then create three agents with FoundryChatClient.as_agent in the Microsoft Agent Framework SDK for Python, wire them with SequentialBuilder(participants=[...], output_from="all"), and print every agent’s message from result.get_outputs(). Watch how each agent builds on the conversation before it, and try other feedback to see whether the classifier’s label changes the recommended action. The exercise notes that some of its technologies are in preview.

How it fits

This is 2.4.1’s ordered pipeline, but nothing checks a stage’s result before the next stage consumes it. Ask where 2.4.1’s gate would go and whether the action agent needs the whole conversation or only the classifier’s response, then compare the prebuilt builder with 2.4.11’s typed executor graph.